<s>
Optical	B-General_Concept
braille	I-General_Concept
recognition	I-General_Concept
is	O
technology	O
to	O
capture	O
and	O
process	O
images	O
of	O
braille	O
characters	O
into	O
natural	O
language	O
characters	O
.	O
</s>
<s>
The	O
system	O
photographed	O
the	O
braille	O
text	O
with	O
a	O
low-resolution	O
CCD	B-Algorithm
camera	I-Algorithm
,	O
and	O
used	O
spatial	O
filtering	O
techniques	O
,	O
median	B-Algorithm
filtering	I-Algorithm
,	O
erosion	B-Algorithm
,	O
and	O
dilation	B-Algorithm
to	O
extract	O
the	O
braille	O
.	O
</s>
<s>
In	O
1993	O
,	O
a	O
group	O
of	O
researchers	O
from	O
the	O
Katholieke	O
Universiteit	O
Leuven	O
developed	O
a	O
system	O
to	O
recognize	O
braille	O
that	O
had	O
been	O
scanned	O
with	O
a	O
commercially	O
available	O
scanner	B-Application
.	O
</s>
<s>
In	O
1999	O
,	O
a	O
group	O
at	O
the	O
Hong	O
Kong	O
Polytechnic	O
University	O
implemented	O
an	O
optical	B-General_Concept
braille	I-General_Concept
recognition	I-General_Concept
technique	O
using	O
edge	B-Algorithm
detection	I-Algorithm
to	O
translate	O
braille	O
into	O
English	O
or	O
Chinese	O
text	O
.	O
</s>
<s>
In	O
2003	O
,	O
Morgavi	O
and	O
Morando	O
designed	O
a	O
system	O
to	O
recognize	O
braille	O
characters	O
using	O
artificial	B-Architecture
neural	I-Architecture
networks	I-Architecture
.	O
</s>
<s>
Some	O
optical	B-General_Concept
braille	I-General_Concept
recognition	I-General_Concept
techniques	O
attempt	O
to	O
use	O
oblique	O
lighting	O
and	O
a	O
camera	O
to	O
reveal	O
the	O
shadows	O
of	O
the	O
depressions	O
and	O
protrusions	O
of	O
the	O
braille	O
.	O
</s>
